Re: Help with corrupted MDADM Raid6

[NeilBrown <neilb@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 1806 bytes --]

On Sat, 14 Jun 2014 11:54:58 +0200 "ptschack ." <ptschack@googlemail.com>
wrote:

> Hi Neil,
> 
> I ran mdadm --examine on the devices (as opposed to the partitions)
> and was surprised:
> Apparently the superblocks are present on some the devices, even if I
> remember differently.
> I attached the output of mdadm to this mail.
> 
> For the sake of clarity, this is how the raid looked like before it
> all went wrong:
> 
> 11 Drives (/dev/sd[abcdefghijk]) alltogether as a RAID 6.
> 2 of those are spares (/dev/sd[jk]).
> 1 is either close to failing of has failed (/dev/sdg).
> 
> I tried
> 
> mdadm --assemble --run /dev/md0 -v /dev/sd[abcdefghi]
> 
> which gave me
> 
> mdadm: looking for devices for /dev/md0
> mdadm: no RAID superblock on /dev/sdg
> mdadm: /dev/sdg has no superblock - assembly aborted
> 
> So it seems I somehow have to restore the superblocks on drives
> /dev/sd[ghijk], or at least on /dev/sd[ghi].
> Is this possible? Any help would be greatly appreciated!
> 
> Regards,
> -P.

Well, you've definitely made progress.  You've found 6 of the devices.
They all look consistent and it appears the array was completely coherent at 
    Mon Jun  9 21:52:48 2014

You think that the 7th device is dead or dying, so you just need to find 2
more (1 would do).

Presumably these are sdh and shi, but it is very strange that we cannot find
the superblock on either of them.
When was the last time the machine was rebooted prio to the date given  -9th
Jun?
Do you have boot logs from that time?  What lines contain 'md'??
Particularly "bind" lines will show you exactly which devices were included.

Maybe also try

  od -x /dev/sdh | grep '4efc a92b'

If the superblock is at some strange location, that might find it.

NeilBrown


[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 828 bytes --]